Terminal value in a pre-revenue technology model is almost always the largest single number in the valuation. In a typical Series A or Series B exercise it accounts for somewhere between seventy and ninety per cent of the present value. That sensitivity is well understood; what is less often acknowledged is how poorly the standard machinery handles the question.
The Gordon growth construction (the perpetuity of a normalised cash flow growing at a constant rate, discounted at the weighted average cost of capital) was developed for businesses whose steady state is observable. Applied to a company that has not yet sold its first unit at scale, it is doing something quite different: it is asking the analyst to specify the steady-state economics of a business that does not exist. The result is a single number that looks precise and is, in fact, an opinion dressed up in arithmetic.
Why the construction misleads
Three features of the pre-revenue case make the perpetuity unreliable.
First, the cash flow being capitalised is a construct. In a mature business, the terminal-year free cash flow is anchored to current performance and adjusted for normalised working capital and maintenance capex. In a pre-revenue case, the analyst is projecting a margin structure that has never been observed in the subject business. Small changes in assumed steady-state margin (three or four percentage points either way) change the valuation by an order of magnitude that is not visible in the model output.
Second, the discount rate is doing work it was not designed to do. The weighted average cost of capital, properly constructed, reflects the risk of the steady-state business. A pre-revenue company carries technology risk, regulatory risk, and competitive-displacement risk that are not steady-state at all; they resolve, in one direction or the other, well before the terminal period begins. Loading those risks into the discount rate produces either an absurdly high cost of capital (which then makes the model insensitive to almost everything else) or a sanitised figure that understates the actual risk.
Third, the growth rate is bounded by macroeconomic logic in a way that is at odds with the actual economics of technology. A constant nominal growth rate above the long-run growth of the economy is incoherent in perpetuity. A constant nominal growth rate at or below that figure understates the structural advantage of a category-defining business in its first decade. Neither answer is right.
What better practice looks like
The institutional approach to this problem is to stop treating the terminal value as a residual and to model an explicit exit. Three constructions deserve more use than they get.
An exit multiple anchored to comparable transactions
Rather than capitalising a normalised cash flow, the analyst projects the business to a defined exit year (commonly five to seven years out) and applies a forward revenue or EBITDA multiple drawn from comparable transactions. The multiple is then itself the subject of sensitivity testing. This produces a terminal value that is grounded in observable market data rather than constructed from assumed steady-state economics.
The criticism of this approach (that it imports current market sentiment into a long-dated valuation) is reasonable and should be addressed by widening the multiple range and showing the committee the full distribution rather than a point estimate.
A staged-perpetuity construction
Where the analyst is committed to a discounted cash flow framework, the perpetuity can be deferred. Instead of capitalising the terminal-year cash flow, the model runs an explicit ten-to-fifteen-year forecast through a maturity transition and only then capitalises a properly normalised steady-state figure. The transition years carry the burden of the actual scaling assumptions; the terminal calculation does what it was designed to do.
This is more demanding to build and harder to explain, but it produces a defensible answer in cases where an exit multiple is not available (a long-horizon infrastructure-adjacent technology, for example, or a vertically integrated business with no clear comparable set).
A probability-weighted exit tree
For binary or trinary outcomes (a drug candidate, a regulatory licence, a platform that either wins its category or does not), the appropriate construction is not a single terminal value at all. It is a small set of outcome states, each with its own exit value and probability, summed under the same discount rate. Investment committees often prefer this output because it makes the underlying bet visible rather than burying it in a single point estimate.
The committee question
One test is worth applying to any pre-revenue valuation. Ask the analyst to show the proportion of present value attributable to cash flows after year ten. If that proportion exceeds two thirds (it usually does), ask what the model is actually saying about the business and what evidence supports that view. The answer is almost never the discount rate. It is the steady-state operating margin and the assumed reinvestment rate, and those are the inputs that deserve the committee's attention.
Terminal value is the most consequential number in the model and, on current practice, it is the number that receives the least scrutiny. The construction can be defended; the convention often cannot.
The perpetuity tells the committee what the analyst believes about a business that has not yet existed. It is a useful disclosure if it is read that way; it is a poor anchor if it is not.
A note on use
None of this argues for abandoning discounted cash flow analysis in venture or growth-stage contexts. It argues for separating the part of the valuation that rests on near-term operating evidence from the part that rests on long-dated assumptions, and for being explicit about which is which. The committee that sees both, in their own terms, makes a better decision than the committee that sees a single weighted average.